Immense random colocalization, revealed by automated high content image cytometry, seriously questions FISH as gold...

EML4-ALK gene fusion (inv2(p21p23)) of non-small cell lung cancer (NSCLC) predisposes to tyrosine kinase inhibitor treatment. One of the gold standard diagnostics is the dual color (DC) break-apart (BA) FISH technique, however, the unusual closeness of the involved genes has been suggested to raise likelihood of random co-localization (RCL) of signals. Although this is suspected to decrease sensitivity (often to as low as 40-70%), the exact level and effect of RCL has not been revealed thus far. Signal distances were analyzed to the 0.1 µm precision in more than 25,000 nuclei, via automated high content-image cytometry. Negative and positive controls were created using conventional DC BA-, and inv2(p21p23) mimicking probe-sets, respectively. Average distance between red and green signals was 9.72 pixels (px) (±5.14px) and 3.28px (±2.44px), in positives and negatives, respectively; overlap in distribution being 41%. Specificity and sensitivity of correctly determining ALK status was 97% and 29%, respectively. When investigating inv2(p21p23) with DC BA FISH, specificity is high, but seven out of ten aberrant nuclei are inevitably falsely classified as negative, due to the extreme level of RCL. Together with genetic heterogeneity and dilution effect of non-tumor cells in NSCLC, this immense analytical false negativity is the primary cause behind the often described low diagnostic sensitivity. These results convincingly suggest that if FISH is to remain a gold standard for detecting the therapy relevant inv(2), either a modified evaluation protocol, or a more reliable probe-design should be considered than the current DC BA one. © 2018 International Society for Advancement of Cytometry.